Each year, as NBA teams make their way towards their goal of winning the championship, new players emerge as superstars. This year has been no different, with a new crop of breakout stars this year. On the other hand, some players fail to perform at their expected level and contribute to their team’s earlier than anticipated exit. Once again, this year has had its fair share of those who have underperformed, hurting their teams’ shots at a championship with their play. Using confidence levels and hypothesis tests, we can find the biggest breakout players and unexpected successes, along with the most disappointing under-performers of the playoffs.
The way that I went about finding these surprising and disappointing players were with hypothesis tests. To conduct these tests, I needed to use one stat that could summarize a player’s contribution to his team and is available on a per game basis. While I could have gone with the simple route and chose to use points, I opted to use box plus-minus (BPM) instead, a rate stat that summarizes both offensive and defensive contributions in addition to being resistant to volume dependent variables like points are.
After gathering the regular season and playoff game logs of many players, each of whom played at least 20 games during the regular season and 3 games in the playoffs, I recorded the mean and standard deviation of their box plus-minus’s in each game. With these summary statistics, I was then able to conduct a one-sided, two-sample t-test, where the alternative hypothesis was that the player’s mean playoff BPM was greater than their mean regular season BPM. Since I wanted to do the calculations myself, I chose to use the conservative method for finding the degrees of freedom. Surprisingly, only 2 players had results that would reject the null hypothesis at a 5% significance level, whereas all the other 73 players failed to reject the null hypothesis. This was likely due to the standard error being so large as all the players played between 3 and 17 playoff games, a very sample size.
In order to present my findings, I took the ten players with the lowest p-values, who were the biggest surprises, and the ten players with the highest p-values, who were the biggest disappointments, and made graphs for each. One of the graphs is a bar graph showing the p-value for the t-test, while the other graph shows a 90% confidence interval (90% because it corresponds with a 5% significance level for a one-sided test) for the difference between mean playoff BPM and mean regular season BPM. Lastly, I included two final graphs of the 90% confidence intervals for all 75 players that fulfilled the criteria previously listed and I found the p-values for.
The Most Surprising Players
By using the difference between each player’s mean playoff and regular season box plus minus, I was able to find the most surprising players. These players have improved greatly over these playoffs, which should help with their development and also future deals they make. Moreover, several young players who improved greatly in the playoffs, such as Jayson Tatum in 2018 and Pascal Siakam in 2019, continued their improvement during the next season.
The p-values presented above should be interpreted as the chance that a player would do better than they actually did assuming their true mean BPM is their regular season BPM. The confidence intervals should be interpreted as the interval for which there is a 90% chance the player’s real playoff BPM falls within.
The most surprising player of the 2020 NBA Playoffs has been Jamal Murray, who has increased his BPM by almost 5 in these playoffs. His outstanding performance has greatly helped the Nuggets reach the conference finals for the first time since 2009. (Note: In my previous article, I stated that the chances of Jamal Murray’s points increasing as much as they did was about 1 in 28000. The differences between that and the results here are that I used BPM instead of points for this study, and I used a one-sample test instead of a two-sample test) The only other statistically significant improvement in the playoffs was Robert Covington of the Houston Rockets, who increased his BPM by just under 5 as well.
While Donovan Mitchell and Mike Conley both saw their BPM’s increase by over 7.5 in the playoffs, their results were not as impressive as Jamal Murray’s and Robert Covington’s. The primary reason for this was sample size, as they played 5 and 7 games, respectively, compared to 17 and 10 games for Murray and Covington. Moreover, Conley displayed wild inconsistency during the playoffs as he had a game with a +27 BPM, which hurt the significance of his results. More interesting results include the improvement of the Heat’s rookie Tyler Herro, which could be a sign of a big sophomore year, and the Celtics duo of Jayson Tatum and Jaylen Brown, signifying the Celtics’ potential for the future.
The Most Disappointing Players
While some superstars are born in the playoffs, others fall. Playoff disappointments like those listed below have destroyed their teams, leaving them to rely on sources who are not as proficient instead. The players that have shown a large decrease in BPM during the postseason will see a fall in value, because they are showing that they cannot play at their best against both the best defenses and in the biggest moments.
The most disappointing player this postseason has been none other than the Toronto Raptors’ Pascal Siakam. He did not play at his regular season level against the Nets in the first round or against the Celtics’ top five defense in the second round, which was a large part of the Raptors’ second round exit. Moreover, Khris Middleton’s regression during the playoffs likely played a large part in their second round exit as well, although most of his poor performances came in the first round against the Orlando Magic.
Two surprising results were Jerami Grant and Paul Millsap, both of whom are on a team that is still in contention for the NBA title with the Nuggets. Both of their BPM’s decreased by just over 2.5 in the playoffs. The reason for Grant’s regression has been shooting and turnovers, as his true shooting percentage has decreased by 3 percentage points, and his turnover percentage has increased by almost 4 percentage points. Paul Millsap suffered from a similar case of shooting woes, with his true shooting percentage falling from 59% in the regular season to 53% in the playoffs.
Two high profile names on this list are Paul George and Russell Westbrook. George’s 3 point percentage went from 41% in the regular season to a below average 33% in the post season. Futhermore, George’s all around performance got worse, as he averaged fewer assists and rebounds per 100 possessions in the playoffs when compared to his regular season averages. Russell Westbrook also shot dreadfully, with a true shooting percentage under 47% in the playoffs. That value was the third worst of this post season among players who shot at least 15 field goals per game, ahead of only Tobias Harris and Pascal Siakam. Even worse, his 46.4% true shooting percentage ranked 7th worst among all stars who shot at least 15 times a game in the last 10 playoffs. The disappointing performances of Paul George and Russell Westbrook hurt their teams greatly.
Jamal Murray and Pascal Siakam were the most surprising and disappointing players, respectively. Murray’s offensive explosion has greatly boosted the Nuggets and improves his long term outlook as a player. Other young players who could blossom into superstars based on their postseason performances include Donovan Mitchell, Tyler Herro, Jayson Tatum, and Jaylen Brown. On the other hand, disappointments like Siakam, Khris Middleton, Paul George, and Russell Westbrook each got worse and hurt their teams’ efforts in the playoffs. While we cannot know whether they were due to random chance or actual regression for sure, it is certainly not a good sign for their future and reputations.
If you are curious about how well your favorite player or some other player not discussed above did in the post season compared to the regular season, all of the confidence intervals for the 75 players who fulfilled the requirements are shown below. They are ranked in ascending order of their p-values, meaning the most statistically significant improvements are at the beginning while the worst performances are at the end.